]>
Commit | Line | Data |
---|---|---|
ddbcc7e8 PM |
1 | #ifndef _LINUX_CGROUP_H |
2 | #define _LINUX_CGROUP_H | |
3 | /* | |
4 | * cgroup interface | |
5 | * | |
6 | * Copyright (C) 2003 BULL SA | |
7 | * Copyright (C) 2004-2006 Silicon Graphics, Inc. | |
8 | * | |
9 | */ | |
10 | ||
11 | #include <linux/sched.h> | |
ddbcc7e8 PM |
12 | #include <linux/cpumask.h> |
13 | #include <linux/nodemask.h> | |
14 | #include <linux/rcupdate.h> | |
846c7bb0 | 15 | #include <linux/cgroupstats.h> |
31a7df01 | 16 | #include <linux/prio_heap.h> |
cc31edce | 17 | #include <linux/rwsem.h> |
38460b48 | 18 | #include <linux/idr.h> |
ddbcc7e8 PM |
19 | |
20 | #ifdef CONFIG_CGROUPS | |
21 | ||
22 | struct cgroupfs_root; | |
23 | struct cgroup_subsys; | |
24 | struct inode; | |
84eea842 | 25 | struct cgroup; |
38460b48 | 26 | struct css_id; |
ddbcc7e8 PM |
27 | |
28 | extern int cgroup_init_early(void); | |
29 | extern int cgroup_init(void); | |
ddbcc7e8 | 30 | extern void cgroup_lock(void); |
d11c563d | 31 | extern int cgroup_lock_is_held(void); |
84eea842 | 32 | extern bool cgroup_lock_live_group(struct cgroup *cgrp); |
ddbcc7e8 | 33 | extern void cgroup_unlock(void); |
b4f48b63 PM |
34 | extern void cgroup_fork(struct task_struct *p); |
35 | extern void cgroup_fork_callbacks(struct task_struct *p); | |
817929ec | 36 | extern void cgroup_post_fork(struct task_struct *p); |
b4f48b63 | 37 | extern void cgroup_exit(struct task_struct *p, int run_callbacks); |
846c7bb0 BS |
38 | extern int cgroupstats_build(struct cgroupstats *stats, |
39 | struct dentry *dentry); | |
e6a1105b | 40 | extern int cgroup_load_subsys(struct cgroup_subsys *ss); |
cf5d5941 | 41 | extern void cgroup_unload_subsys(struct cgroup_subsys *ss); |
ddbcc7e8 | 42 | |
828c0950 | 43 | extern const struct file_operations proc_cgroup_operations; |
a424316c | 44 | |
aae8aab4 | 45 | /* Define the enumeration of all builtin cgroup subsystems */ |
817929ec PM |
46 | #define SUBSYS(_x) _x ## _subsys_id, |
47 | enum cgroup_subsys_id { | |
48 | #include <linux/cgroup_subsys.h> | |
aae8aab4 | 49 | CGROUP_BUILTIN_SUBSYS_COUNT |
817929ec PM |
50 | }; |
51 | #undef SUBSYS | |
aae8aab4 BB |
52 | /* |
53 | * This define indicates the maximum number of subsystems that can be loaded | |
54 | * at once. We limit to this many since cgroupfs_root has subsys_bits to keep | |
55 | * track of all of them. | |
56 | */ | |
57 | #define CGROUP_SUBSYS_COUNT (BITS_PER_BYTE*sizeof(unsigned long)) | |
817929ec | 58 | |
ddbcc7e8 PM |
59 | /* Per-subsystem/per-cgroup state maintained by the system. */ |
60 | struct cgroup_subsys_state { | |
d20a390a PM |
61 | /* |
62 | * The cgroup that this subsystem is attached to. Useful | |
ddbcc7e8 | 63 | * for subsystems that want to know about the cgroup |
d20a390a PM |
64 | * hierarchy structure |
65 | */ | |
ddbcc7e8 PM |
66 | struct cgroup *cgroup; |
67 | ||
d20a390a PM |
68 | /* |
69 | * State maintained by the cgroup system to allow subsystems | |
e7c5ec91 | 70 | * to be "busy". Should be accessed via css_get(), |
d20a390a PM |
71 | * css_tryget() and and css_put(). |
72 | */ | |
ddbcc7e8 PM |
73 | |
74 | atomic_t refcnt; | |
75 | ||
76 | unsigned long flags; | |
38460b48 | 77 | /* ID for this css, if possible */ |
2c392b8c | 78 | struct css_id __rcu *id; |
ddbcc7e8 PM |
79 | }; |
80 | ||
81 | /* bits in struct cgroup_subsys_state flags field */ | |
82 | enum { | |
83 | CSS_ROOT, /* This CSS is the root of the subsystem */ | |
e7c5ec91 | 84 | CSS_REMOVED, /* This CSS is dead */ |
ddbcc7e8 PM |
85 | }; |
86 | ||
d7b9fff7 DN |
87 | /* Caller must verify that the css is not for root cgroup */ |
88 | static inline void __css_get(struct cgroup_subsys_state *css, int count) | |
89 | { | |
90 | atomic_add(count, &css->refcnt); | |
91 | } | |
92 | ||
ddbcc7e8 | 93 | /* |
e7c5ec91 PM |
94 | * Call css_get() to hold a reference on the css; it can be used |
95 | * for a reference obtained via: | |
96 | * - an existing ref-counted reference to the css | |
97 | * - task->cgroups for a locked task | |
ddbcc7e8 PM |
98 | */ |
99 | ||
100 | static inline void css_get(struct cgroup_subsys_state *css) | |
101 | { | |
102 | /* We don't need to reference count the root state */ | |
103 | if (!test_bit(CSS_ROOT, &css->flags)) | |
d7b9fff7 | 104 | __css_get(css, 1); |
ddbcc7e8 | 105 | } |
e7c5ec91 PM |
106 | |
107 | static inline bool css_is_removed(struct cgroup_subsys_state *css) | |
108 | { | |
109 | return test_bit(CSS_REMOVED, &css->flags); | |
110 | } | |
111 | ||
112 | /* | |
113 | * Call css_tryget() to take a reference on a css if your existing | |
114 | * (known-valid) reference isn't already ref-counted. Returns false if | |
115 | * the css has been destroyed. | |
116 | */ | |
117 | ||
28b4c27b | 118 | extern bool __css_tryget(struct cgroup_subsys_state *css); |
e7c5ec91 PM |
119 | static inline bool css_tryget(struct cgroup_subsys_state *css) |
120 | { | |
121 | if (test_bit(CSS_ROOT, &css->flags)) | |
122 | return true; | |
28b4c27b | 123 | return __css_tryget(css); |
e7c5ec91 PM |
124 | } |
125 | ||
ddbcc7e8 PM |
126 | /* |
127 | * css_put() should be called to release a reference taken by | |
e7c5ec91 | 128 | * css_get() or css_tryget() |
ddbcc7e8 PM |
129 | */ |
130 | ||
28b4c27b | 131 | extern void __css_put(struct cgroup_subsys_state *css); |
ddbcc7e8 PM |
132 | static inline void css_put(struct cgroup_subsys_state *css) |
133 | { | |
134 | if (!test_bit(CSS_ROOT, &css->flags)) | |
28b4c27b | 135 | __css_put(css); |
ddbcc7e8 PM |
136 | } |
137 | ||
3116f0e3 PM |
138 | /* bits in struct cgroup flags field */ |
139 | enum { | |
140 | /* Control Group is dead */ | |
141 | CGRP_REMOVED, | |
d20a390a PM |
142 | /* |
143 | * Control Group has previously had a child cgroup or a task, | |
144 | * but no longer (only if CGRP_NOTIFY_ON_RELEASE is set) | |
145 | */ | |
3116f0e3 PM |
146 | CGRP_RELEASABLE, |
147 | /* Control Group requires release notifications to userspace */ | |
148 | CGRP_NOTIFY_ON_RELEASE, | |
ec64f515 KH |
149 | /* |
150 | * A thread in rmdir() is wating for this cgroup. | |
151 | */ | |
152 | CGRP_WAIT_ON_RMDIR, | |
97978e6d DL |
153 | /* |
154 | * Clone cgroup values when creating a new child cgroup | |
155 | */ | |
156 | CGRP_CLONE_CHILDREN, | |
3116f0e3 PM |
157 | }; |
158 | ||
ddbcc7e8 PM |
159 | struct cgroup { |
160 | unsigned long flags; /* "unsigned long" so bitops work */ | |
161 | ||
d20a390a PM |
162 | /* |
163 | * count users of this cgroup. >0 means busy, but doesn't | |
164 | * necessarily indicate the number of tasks in the cgroup | |
165 | */ | |
ddbcc7e8 PM |
166 | atomic_t count; |
167 | ||
168 | /* | |
169 | * We link our 'sibling' struct into our parent's 'children'. | |
170 | * Our children link their 'sibling' into our 'children'. | |
171 | */ | |
172 | struct list_head sibling; /* my parent's children */ | |
173 | struct list_head children; /* my children */ | |
05ef1d7c | 174 | struct list_head files; /* my files */ |
ddbcc7e8 | 175 | |
d20a390a | 176 | struct cgroup *parent; /* my parent */ |
2c392b8c | 177 | struct dentry __rcu *dentry; /* cgroup fs entry, RCU protected */ |
ddbcc7e8 PM |
178 | |
179 | /* Private pointers for each registered subsystem */ | |
180 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
181 | ||
182 | struct cgroupfs_root *root; | |
183 | struct cgroup *top_cgroup; | |
817929ec PM |
184 | |
185 | /* | |
186 | * List of cg_cgroup_links pointing at css_sets with | |
187 | * tasks in this cgroup. Protected by css_set_lock | |
188 | */ | |
189 | struct list_head css_sets; | |
81a6a5cd | 190 | |
b0ca5a84 | 191 | struct list_head allcg_node; /* cgroupfs_root->allcg_list */ |
8e3f6541 | 192 | struct list_head cft_q_node; /* used during cftype add/rm */ |
b0ca5a84 | 193 | |
81a6a5cd PM |
194 | /* |
195 | * Linked list running through all cgroups that can | |
196 | * potentially be reaped by the release agent. Protected by | |
197 | * release_list_lock | |
198 | */ | |
199 | struct list_head release_list; | |
cc31edce | 200 | |
72a8cb30 BB |
201 | /* |
202 | * list of pidlists, up to two for each namespace (one for procs, one | |
203 | * for tasks); created on demand. | |
204 | */ | |
205 | struct list_head pidlists; | |
206 | struct mutex pidlist_mutex; | |
a47295e6 PM |
207 | |
208 | /* For RCU-protected deletion */ | |
209 | struct rcu_head rcu_head; | |
0dea1168 | 210 | |
25985edc | 211 | /* List of events which userspace want to receive */ |
0dea1168 KS |
212 | struct list_head event_list; |
213 | spinlock_t event_list_lock; | |
817929ec PM |
214 | }; |
215 | ||
d20a390a PM |
216 | /* |
217 | * A css_set is a structure holding pointers to a set of | |
817929ec PM |
218 | * cgroup_subsys_state objects. This saves space in the task struct |
219 | * object and speeds up fork()/exit(), since a single inc/dec and a | |
d20a390a PM |
220 | * list_add()/del() can bump the reference count on the entire cgroup |
221 | * set for a task. | |
817929ec PM |
222 | */ |
223 | ||
224 | struct css_set { | |
225 | ||
226 | /* Reference count */ | |
146aa1bd | 227 | atomic_t refcount; |
817929ec | 228 | |
472b1053 LZ |
229 | /* |
230 | * List running through all cgroup groups in the same hash | |
231 | * slot. Protected by css_set_lock | |
232 | */ | |
233 | struct hlist_node hlist; | |
234 | ||
817929ec PM |
235 | /* |
236 | * List running through all tasks using this cgroup | |
237 | * group. Protected by css_set_lock | |
238 | */ | |
239 | struct list_head tasks; | |
240 | ||
241 | /* | |
242 | * List of cg_cgroup_link objects on link chains from | |
243 | * cgroups referenced from this css_set. Protected by | |
244 | * css_set_lock | |
245 | */ | |
246 | struct list_head cg_links; | |
247 | ||
248 | /* | |
249 | * Set of subsystem states, one for each subsystem. This array | |
250 | * is immutable after creation apart from the init_css_set | |
cf5d5941 BB |
251 | * during subsystem registration (at boot time) and modular subsystem |
252 | * loading/unloading. | |
817929ec PM |
253 | */ |
254 | struct cgroup_subsys_state *subsys[CGROUP_SUBSYS_COUNT]; | |
c378369d BB |
255 | |
256 | /* For RCU-protected deletion */ | |
257 | struct rcu_head rcu_head; | |
ddbcc7e8 PM |
258 | }; |
259 | ||
91796569 PM |
260 | /* |
261 | * cgroup_map_cb is an abstract callback API for reporting map-valued | |
262 | * control files | |
263 | */ | |
264 | ||
265 | struct cgroup_map_cb { | |
266 | int (*fill)(struct cgroup_map_cb *cb, const char *key, u64 value); | |
267 | void *state; | |
268 | }; | |
269 | ||
d20a390a PM |
270 | /* |
271 | * struct cftype: handler definitions for cgroup control files | |
ddbcc7e8 PM |
272 | * |
273 | * When reading/writing to a file: | |
a043e3b2 | 274 | * - the cgroup to use is file->f_dentry->d_parent->d_fsdata |
ddbcc7e8 PM |
275 | * - the 'cftype' of the file is file->f_dentry->d_fsdata |
276 | */ | |
277 | ||
8e3f6541 TH |
278 | /* cftype->flags */ |
279 | #define CFTYPE_ONLY_ON_ROOT (1U << 0) /* only create on root cg */ | |
280 | #define CFTYPE_NOT_ON_ROOT (1U << 1) /* don't create onp root cg */ | |
281 | ||
282 | #define MAX_CFTYPE_NAME 64 | |
283 | ||
ddbcc7e8 | 284 | struct cftype { |
d20a390a PM |
285 | /* |
286 | * By convention, the name should begin with the name of the | |
8e3f6541 TH |
287 | * subsystem, followed by a period. Zero length string indicates |
288 | * end of cftype array. | |
d20a390a | 289 | */ |
ddbcc7e8 PM |
290 | char name[MAX_CFTYPE_NAME]; |
291 | int private; | |
099fca32 LZ |
292 | /* |
293 | * If not 0, file mode is set to this value, otherwise it will | |
294 | * be figured out automatically | |
295 | */ | |
a5e7ed32 | 296 | umode_t mode; |
db3b1497 PM |
297 | |
298 | /* | |
299 | * If non-zero, defines the maximum length of string that can | |
300 | * be passed to write_string; defaults to 64 | |
301 | */ | |
302 | size_t max_write_len; | |
303 | ||
8e3f6541 TH |
304 | /* CFTYPE_* flags */ |
305 | unsigned int flags; | |
306 | ||
ce16b49d PM |
307 | int (*open)(struct inode *inode, struct file *file); |
308 | ssize_t (*read)(struct cgroup *cgrp, struct cftype *cft, | |
309 | struct file *file, | |
310 | char __user *buf, size_t nbytes, loff_t *ppos); | |
ddbcc7e8 | 311 | /* |
f4c753b7 | 312 | * read_u64() is a shortcut for the common case of returning a |
ddbcc7e8 PM |
313 | * single integer. Use it in place of read() |
314 | */ | |
ce16b49d | 315 | u64 (*read_u64)(struct cgroup *cgrp, struct cftype *cft); |
e73d2c61 PM |
316 | /* |
317 | * read_s64() is a signed version of read_u64() | |
318 | */ | |
ce16b49d | 319 | s64 (*read_s64)(struct cgroup *cgrp, struct cftype *cft); |
91796569 PM |
320 | /* |
321 | * read_map() is used for defining a map of key/value | |
322 | * pairs. It should call cb->fill(cb, key, value) for each | |
323 | * entry. The key/value pairs (and their ordering) should not | |
324 | * change between reboots. | |
325 | */ | |
ce16b49d PM |
326 | int (*read_map)(struct cgroup *cont, struct cftype *cft, |
327 | struct cgroup_map_cb *cb); | |
29486df3 SH |
328 | /* |
329 | * read_seq_string() is used for outputting a simple sequence | |
330 | * using seqfile. | |
331 | */ | |
ce16b49d PM |
332 | int (*read_seq_string)(struct cgroup *cont, struct cftype *cft, |
333 | struct seq_file *m); | |
91796569 | 334 | |
ce16b49d PM |
335 | ssize_t (*write)(struct cgroup *cgrp, struct cftype *cft, |
336 | struct file *file, | |
337 | const char __user *buf, size_t nbytes, loff_t *ppos); | |
355e0c48 PM |
338 | |
339 | /* | |
f4c753b7 | 340 | * write_u64() is a shortcut for the common case of accepting |
355e0c48 PM |
341 | * a single integer (as parsed by simple_strtoull) from |
342 | * userspace. Use in place of write(); return 0 or error. | |
343 | */ | |
ce16b49d | 344 | int (*write_u64)(struct cgroup *cgrp, struct cftype *cft, u64 val); |
e73d2c61 PM |
345 | /* |
346 | * write_s64() is a signed version of write_u64() | |
347 | */ | |
ce16b49d | 348 | int (*write_s64)(struct cgroup *cgrp, struct cftype *cft, s64 val); |
355e0c48 | 349 | |
db3b1497 PM |
350 | /* |
351 | * write_string() is passed a nul-terminated kernelspace | |
352 | * buffer of maximum length determined by max_write_len. | |
353 | * Returns 0 or -ve error code. | |
354 | */ | |
355 | int (*write_string)(struct cgroup *cgrp, struct cftype *cft, | |
356 | const char *buffer); | |
d447ea2f PE |
357 | /* |
358 | * trigger() callback can be used to get some kick from the | |
359 | * userspace, when the actual string written is not important | |
360 | * at all. The private field can be used to determine the | |
361 | * kick type for multiplexing. | |
362 | */ | |
363 | int (*trigger)(struct cgroup *cgrp, unsigned int event); | |
364 | ||
ce16b49d | 365 | int (*release)(struct inode *inode, struct file *file); |
0dea1168 KS |
366 | |
367 | /* | |
368 | * register_event() callback will be used to add new userspace | |
369 | * waiter for changes related to the cftype. Implement it if | |
370 | * you want to provide this functionality. Use eventfd_signal() | |
371 | * on eventfd to send notification to userspace. | |
372 | */ | |
373 | int (*register_event)(struct cgroup *cgrp, struct cftype *cft, | |
374 | struct eventfd_ctx *eventfd, const char *args); | |
375 | /* | |
376 | * unregister_event() callback will be called when userspace | |
377 | * closes the eventfd or on cgroup removing. | |
378 | * This callback must be implemented, if you want provide | |
379 | * notification functionality. | |
0dea1168 | 380 | */ |
907860ed | 381 | void (*unregister_event)(struct cgroup *cgrp, struct cftype *cft, |
0dea1168 | 382 | struct eventfd_ctx *eventfd); |
ddbcc7e8 PM |
383 | }; |
384 | ||
8e3f6541 TH |
385 | /* |
386 | * cftype_sets describe cftypes belonging to a subsystem and are chained at | |
387 | * cgroup_subsys->cftsets. Each cftset points to an array of cftypes | |
388 | * terminated by zero length name. | |
389 | */ | |
390 | struct cftype_set { | |
391 | struct list_head node; /* chained at subsys->cftsets */ | |
392 | const struct cftype *cfts; | |
393 | }; | |
394 | ||
31a7df01 CW |
395 | struct cgroup_scanner { |
396 | struct cgroup *cg; | |
397 | int (*test_task)(struct task_struct *p, struct cgroup_scanner *scan); | |
398 | void (*process_task)(struct task_struct *p, | |
399 | struct cgroup_scanner *scan); | |
400 | struct ptr_heap *heap; | |
bd1a8ab7 | 401 | void *data; |
31a7df01 CW |
402 | }; |
403 | ||
8e3f6541 | 404 | int cgroup_add_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts); |
79578621 | 405 | int cgroup_rm_cftypes(struct cgroup_subsys *ss, const struct cftype *cfts); |
8e3f6541 | 406 | |
ffd2d883 | 407 | int cgroup_is_removed(const struct cgroup *cgrp); |
ddbcc7e8 | 408 | |
ffd2d883 | 409 | int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen); |
ddbcc7e8 | 410 | |
ffd2d883 | 411 | int cgroup_task_count(const struct cgroup *cgrp); |
bbcb81d0 | 412 | |
313e924c GN |
413 | /* Return true if cgrp is a descendant of the task's cgroup */ |
414 | int cgroup_is_descendant(const struct cgroup *cgrp, struct task_struct *task); | |
ddbcc7e8 | 415 | |
88703267 KH |
416 | /* |
417 | * When the subsys has to access css and may add permanent refcnt to css, | |
418 | * it should take care of racy conditions with rmdir(). Following set of | |
419 | * functions, is for stop/restart rmdir if necessary. | |
420 | * Because these will call css_get/put, "css" should be alive css. | |
421 | * | |
422 | * cgroup_exclude_rmdir(); | |
423 | * ...do some jobs which may access arbitrary empty cgroup | |
424 | * cgroup_release_and_wakeup_rmdir(); | |
425 | * | |
426 | * When someone removes a cgroup while cgroup_exclude_rmdir() holds it, | |
427 | * it sleeps and cgroup_release_and_wakeup_rmdir() will wake him up. | |
428 | */ | |
429 | ||
430 | void cgroup_exclude_rmdir(struct cgroup_subsys_state *css); | |
431 | void cgroup_release_and_wakeup_rmdir(struct cgroup_subsys_state *css); | |
432 | ||
2f7ee569 TH |
433 | /* |
434 | * Control Group taskset, used to pass around set of tasks to cgroup_subsys | |
435 | * methods. | |
436 | */ | |
437 | struct cgroup_taskset; | |
438 | struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset); | |
439 | struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset); | |
440 | struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset); | |
441 | int cgroup_taskset_size(struct cgroup_taskset *tset); | |
442 | ||
443 | /** | |
444 | * cgroup_taskset_for_each - iterate cgroup_taskset | |
445 | * @task: the loop cursor | |
446 | * @skip_cgrp: skip if task's cgroup matches this, %NULL to iterate through all | |
447 | * @tset: taskset to iterate | |
448 | */ | |
449 | #define cgroup_taskset_for_each(task, skip_cgrp, tset) \ | |
450 | for ((task) = cgroup_taskset_first((tset)); (task); \ | |
451 | (task) = cgroup_taskset_next((tset))) \ | |
452 | if (!(skip_cgrp) || \ | |
453 | cgroup_taskset_cur_cgroup((tset)) != (skip_cgrp)) | |
454 | ||
21acb9ca TLSC |
455 | /* |
456 | * Control Group subsystem type. | |
457 | * See Documentation/cgroups/cgroups.txt for details | |
458 | */ | |
ddbcc7e8 PM |
459 | |
460 | struct cgroup_subsys { | |
761b3ef5 LZ |
461 | struct cgroup_subsys_state *(*create)(struct cgroup *cgrp); |
462 | int (*pre_destroy)(struct cgroup *cgrp); | |
463 | void (*destroy)(struct cgroup *cgrp); | |
464 | int (*can_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
465 | void (*cancel_attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
466 | void (*attach)(struct cgroup *cgrp, struct cgroup_taskset *tset); | |
467 | void (*fork)(struct task_struct *task); | |
468 | void (*exit)(struct cgroup *cgrp, struct cgroup *old_cgrp, | |
469 | struct task_struct *task); | |
470 | int (*populate)(struct cgroup_subsys *ss, struct cgroup *cgrp); | |
471 | void (*post_clone)(struct cgroup *cgrp); | |
472 | void (*bind)(struct cgroup *root); | |
e5991371 | 473 | |
ddbcc7e8 PM |
474 | int subsys_id; |
475 | int active; | |
8bab8dde | 476 | int disabled; |
ddbcc7e8 | 477 | int early_init; |
38460b48 KH |
478 | /* |
479 | * True if this subsys uses ID. ID is not available before cgroup_init() | |
480 | * (not available in early_init time.) | |
481 | */ | |
482 | bool use_id; | |
ddbcc7e8 PM |
483 | #define MAX_CGROUP_TYPE_NAMELEN 32 |
484 | const char *name; | |
485 | ||
999cd8a4 PM |
486 | /* |
487 | * Protects sibling/children links of cgroups in this | |
488 | * hierarchy, plus protects which hierarchy (or none) the | |
489 | * subsystem is a part of (i.e. root/sibling). To avoid | |
490 | * potential deadlocks, the following operations should not be | |
491 | * undertaken while holding any hierarchy_mutex: | |
492 | * | |
493 | * - allocating memory | |
494 | * - initiating hotplug events | |
495 | */ | |
496 | struct mutex hierarchy_mutex; | |
cfebe563 | 497 | struct lock_class_key subsys_key; |
ddbcc7e8 | 498 | |
999cd8a4 PM |
499 | /* |
500 | * Link to parent, and list entry in parent's children. | |
501 | * Protected by this->hierarchy_mutex and cgroup_lock() | |
502 | */ | |
503 | struct cgroupfs_root *root; | |
ddbcc7e8 | 504 | struct list_head sibling; |
38460b48 KH |
505 | /* used when use_id == true */ |
506 | struct idr idr; | |
42aee6c4 | 507 | spinlock_t id_lock; |
e6a1105b | 508 | |
8e3f6541 TH |
509 | /* list of cftype_sets */ |
510 | struct list_head cftsets; | |
511 | ||
512 | /* base cftypes, automatically [de]registered with subsys itself */ | |
513 | struct cftype *base_cftypes; | |
514 | struct cftype_set base_cftset; | |
515 | ||
e6a1105b BB |
516 | /* should be defined only by modular subsystems */ |
517 | struct module *module; | |
ddbcc7e8 PM |
518 | }; |
519 | ||
520 | #define SUBSYS(_x) extern struct cgroup_subsys _x ## _subsys; | |
521 | #include <linux/cgroup_subsys.h> | |
522 | #undef SUBSYS | |
523 | ||
524 | static inline struct cgroup_subsys_state *cgroup_subsys_state( | |
ffd2d883 | 525 | struct cgroup *cgrp, int subsys_id) |
ddbcc7e8 | 526 | { |
ffd2d883 | 527 | return cgrp->subsys[subsys_id]; |
ddbcc7e8 PM |
528 | } |
529 | ||
dc61b1d6 PZ |
530 | /* |
531 | * function to get the cgroup_subsys_state which allows for extra | |
532 | * rcu_dereference_check() conditions, such as locks used during the | |
533 | * cgroup_subsys::attach() methods. | |
534 | */ | |
535 | #define task_subsys_state_check(task, subsys_id, __c) \ | |
536 | rcu_dereference_check(task->cgroups->subsys[subsys_id], \ | |
dc61b1d6 PZ |
537 | lockdep_is_held(&task->alloc_lock) || \ |
538 | cgroup_lock_is_held() || (__c)) | |
539 | ||
540 | static inline struct cgroup_subsys_state * | |
541 | task_subsys_state(struct task_struct *task, int subsys_id) | |
ddbcc7e8 | 542 | { |
dc61b1d6 | 543 | return task_subsys_state_check(task, subsys_id, false); |
ddbcc7e8 PM |
544 | } |
545 | ||
546 | static inline struct cgroup* task_cgroup(struct task_struct *task, | |
547 | int subsys_id) | |
548 | { | |
549 | return task_subsys_state(task, subsys_id)->cgroup; | |
550 | } | |
551 | ||
817929ec PM |
552 | /* A cgroup_iter should be treated as an opaque object */ |
553 | struct cgroup_iter { | |
554 | struct list_head *cg_link; | |
555 | struct list_head *task; | |
556 | }; | |
557 | ||
d20a390a PM |
558 | /* |
559 | * To iterate across the tasks in a cgroup: | |
817929ec | 560 | * |
b595076a | 561 | * 1) call cgroup_iter_start to initialize an iterator |
817929ec PM |
562 | * |
563 | * 2) call cgroup_iter_next() to retrieve member tasks until it | |
564 | * returns NULL or until you want to end the iteration | |
565 | * | |
566 | * 3) call cgroup_iter_end() to destroy the iterator. | |
31a7df01 | 567 | * |
d20a390a PM |
568 | * Or, call cgroup_scan_tasks() to iterate through every task in a |
569 | * cgroup - cgroup_scan_tasks() holds the css_set_lock when calling | |
570 | * the test_task() callback, but not while calling the process_task() | |
571 | * callback. | |
817929ec | 572 | */ |
ffd2d883 LZ |
573 | void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it); |
574 | struct task_struct *cgroup_iter_next(struct cgroup *cgrp, | |
817929ec | 575 | struct cgroup_iter *it); |
ffd2d883 | 576 | void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it); |
31a7df01 | 577 | int cgroup_scan_tasks(struct cgroup_scanner *scan); |
956db3ca | 578 | int cgroup_attach_task(struct cgroup *, struct task_struct *); |
31583bb0 MT |
579 | int cgroup_attach_task_all(struct task_struct *from, struct task_struct *); |
580 | ||
38460b48 KH |
581 | /* |
582 | * CSS ID is ID for cgroup_subsys_state structs under subsys. This only works | |
583 | * if cgroup_subsys.use_id == true. It can be used for looking up and scanning. | |
584 | * CSS ID is assigned at cgroup allocation (create) automatically | |
585 | * and removed when subsys calls free_css_id() function. This is because | |
586 | * the lifetime of cgroup_subsys_state is subsys's matter. | |
587 | * | |
588 | * Looking up and scanning function should be called under rcu_read_lock(). | |
589 | * Taking cgroup_mutex()/hierarchy_mutex() is not necessary for following calls. | |
590 | * But the css returned by this routine can be "not populated yet" or "being | |
591 | * destroyed". The caller should check css and cgroup's status. | |
592 | */ | |
593 | ||
594 | /* | |
595 | * Typically Called at ->destroy(), or somewhere the subsys frees | |
596 | * cgroup_subsys_state. | |
597 | */ | |
598 | void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css); | |
599 | ||
600 | /* Find a cgroup_subsys_state which has given ID */ | |
601 | ||
602 | struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id); | |
603 | ||
604 | /* | |
605 | * Get a cgroup whose id is greater than or equal to id under tree of root. | |
606 | * Returning a cgroup_subsys_state or NULL. | |
607 | */ | |
608 | struct cgroup_subsys_state *css_get_next(struct cgroup_subsys *ss, int id, | |
609 | struct cgroup_subsys_state *root, int *foundid); | |
610 | ||
611 | /* Returns true if root is ancestor of cg */ | |
612 | bool css_is_ancestor(struct cgroup_subsys_state *cg, | |
0b7f569e | 613 | const struct cgroup_subsys_state *root); |
38460b48 KH |
614 | |
615 | /* Get id and depth of css */ | |
616 | unsigned short css_id(struct cgroup_subsys_state *css); | |
617 | unsigned short css_depth(struct cgroup_subsys_state *css); | |
e5d1367f | 618 | struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id); |
38460b48 | 619 | |
ddbcc7e8 PM |
620 | #else /* !CONFIG_CGROUPS */ |
621 | ||
622 | static inline int cgroup_init_early(void) { return 0; } | |
623 | static inline int cgroup_init(void) { return 0; } | |
b4f48b63 PM |
624 | static inline void cgroup_fork(struct task_struct *p) {} |
625 | static inline void cgroup_fork_callbacks(struct task_struct *p) {} | |
817929ec | 626 | static inline void cgroup_post_fork(struct task_struct *p) {} |
b4f48b63 | 627 | static inline void cgroup_exit(struct task_struct *p, int callbacks) {} |
ddbcc7e8 PM |
628 | |
629 | static inline void cgroup_lock(void) {} | |
630 | static inline void cgroup_unlock(void) {} | |
846c7bb0 BS |
631 | static inline int cgroupstats_build(struct cgroupstats *stats, |
632 | struct dentry *dentry) | |
633 | { | |
634 | return -EINVAL; | |
635 | } | |
ddbcc7e8 | 636 | |
d7926ee3 | 637 | /* No cgroups - nothing to do */ |
31583bb0 MT |
638 | static inline int cgroup_attach_task_all(struct task_struct *from, |
639 | struct task_struct *t) | |
640 | { | |
641 | return 0; | |
642 | } | |
d7926ee3 | 643 | |
ddbcc7e8 PM |
644 | #endif /* !CONFIG_CGROUPS */ |
645 | ||
646 | #endif /* _LINUX_CGROUP_H */ |